In [7]:
import matplotlib.pyplot as plt
plt.figure(figsize=(15, 15))
pos = nx.drawing.layout.bipartite_layout(B, primary_col)
nx.draw_networkx(B, pos=pos, node_size=150, width=0.2, node_color='red')
import matplotlib.pyplot as plt
plt.style.use('seaborn')
import pandas as pd
PRIMARY = 'Category Name'
SECONDARY = 'Order Region'
df = pd.read_csv(
'DataCoSupplyChainDataset.csv',
usecols = [PRIMARY, SECONDARY],
encoding = 'unicode_escape'
).apply(lambda col: col.str.strip())
df.head()
| Category Name | Order Region | |
|---|---|---|
| 0 | Sporting Goods | Southeast Asia |
| 1 | Sporting Goods | South Asia |
| 2 | Sporting Goods | South Asia |
| 3 | Sporting Goods | Oceania |
| 4 | Sporting Goods | Oceania |
df.isnull().sum().sort_values(ascending=False)
Category Name 0 Order Region 0 dtype: int64
df.drop_duplicates(inplace=True)
df.info()
<class 'pandas.core.frame.DataFrame'> Int64Index: 691 entries, 0 to 162002 Data columns (total 2 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Category Name 691 non-null object 1 Order Region 691 non-null object dtypes: object(2) memory usage: 16.2+ KB
primary_col, secondary_col = df[PRIMARY], df[SECONDARY]
print(f'Number of {PRIMARY}:', primary_col.nunique())
print(f'Number of {SECONDARY}:', secondary_col.nunique())
print('Number of edges:', len(df))
Number of Category Name: 50 Number of Order Region: 23 Number of edges: 691
import networkx as nx
from networkx.algorithms import bipartite
B = nx.Graph()
for index, row in df.iterrows():
B.add_edge(row[PRIMARY], row[SECONDARY], weight=1)
B.add_nodes_from(primary_col, bipartite=0)
B.add_nodes_from(secondary_col, bipartite=1)
import matplotlib.pyplot as plt
plt.figure(figsize=(15, 15))
pos = nx.drawing.layout.bipartite_layout(B, primary_col)
nx.draw_networkx(B, pos=pos, node_size=150, width=0.2, node_color='red')
G = bipartite.weighted_projected_graph(B, primary_col)
node_labels = dict(zip(primary_col, primary_col))
plt.figure(figsize=(15, 7))
pos = nx.spring_layout(G)
nx.draw_networkx_nodes(G, pos, nodelist=primary_col, node_color='yellow')
nx.draw_networkx_edges(G, pos, edge_color='grey', alpha=0.2)
nx.draw_networkx_labels(G, pos, labels=node_labels)
plt.show()
df_groupby_category = df.groupby([SECONDARY])[PRIMARY].apply(lambda x: sorted(set(x)))
df_groupby_category = df_groupby_category.reset_index()
# print(df_groupby_category.iloc[14][1])
df_groupby_category.head()
| Order Region | Category Name | |
|---|---|---|
| 0 | Canada | [Accessories, Baseball & Softball, Boxing & MM... |
| 1 | Caribbean | [Accessories, As Seen on TV!, Baseball & Soft... |
| 2 | Central Africa | [Accessories, Baseball & Softball, Boxing & MM... |
| 3 | Central America | [Accessories, As Seen on TV!, Baseball & Soft... |
| 4 | Central Asia | [Accessories, Boxing & MMA, Camping & Hiking, ... |
def get_cluster_common_values(cluster):
commons = []
for index, row in df_groupby_category.iterrows():
if set(cluster).issubset(row[PRIMARY]):
commons.append(row[SECONDARY])
return commons
def print_communities(node_groups):
print('Number of communities:', len(node_groups))
for index, cluster in enumerate(node_groups):
cluster = sorted(cluster)
common_values = sorted(get_cluster_common_values(cluster))
print(f'\nCluster {index}:')
print(f"- {len(cluster)} Nodes: {', '.join(cluster)}")
print(f"- {len(common_values)} Common values: {', '.join(common_values)}")
import matplotlib.cm as cm
def plot_communities(nodes, labels):
cmap = cm.get_cmap('autumn', max(labels) + 1)
pos = nx.spring_layout(G)
plt.figure(figsize=(15, 7))
nx.draw_networkx_nodes(G, pos, nodes, cmap=cmap, node_color=labels)
nx.draw_networkx_edges(G, pos, edge_color='grey', alpha=0.2)
nx.draw_networkx_labels(G, pos)
plt.show()
import community.community_louvain as community_louvain
partition = community_louvain.best_partition(G)
louvain_node_groups = [[] for _ in set(partition.values())]
for node, cluster in sorted(partition.items()):
louvain_node_groups[cluster].append(node)
print_communities(louvain_node_groups)
plot_communities(partition.keys(), list(partition.values()))
Number of communities: 3 Cluster 0: - 24 Nodes: Accessories, Baseball & Softball, Boxing & MMA, Camping & Hiking, Cardio Equipment, Cleats, Electronics, Fishing, Fitness Accessories, Girls' Apparel, Golf Apparel, Golf Balls, Golf Gloves, Golf Shoes, Hockey, Hunting & Shooting, Indoor/Outdoor Games, Lacrosse, Men's Footwear, Shop By Sport, Tennis & Racquet, Trade-In, Water Sports, Women's Apparel - 22 Common values: Canada, Caribbean, Central Africa, Central America, East Africa, East of USA, Eastern Asia, Eastern Europe, North Africa, Northern Europe, Oceania, South America, South Asia, South of USA, Southeast Asia, Southern Africa, Southern Europe, US Center, West Africa, West Asia, West of USA, Western Europe Cluster 1: - 18 Nodes: Baby, Books, CDs, Cameras, Children's Clothing, Computers, Consumer Electronics, Crafts, DVDs, Garden, Health and Beauty, Men's Clothing, Music, Pet Supplies, Sporting Goods, Toys, Video Games, Women's Clothing - 3 Common values: Oceania, South Asia, Southeast Asia Cluster 2: - 8 Nodes: As Seen on TV!, Basketball, Golf Bags & Carts, Kids' Golf Clubs, Men's Golf Clubs, Soccer, Strength Training, Women's Golf Clubs - 3 Common values: Northern Europe, Southern Europe, Western Europe
from sklearn.cluster import KMeans
from scipy.spatial.distance import cdist
adj_matrix = nx.to_numpy_array(G)
adj_matrix
array([[ 0., 4., 4., ..., 4., 4., 4.],
[ 4., 0., 23., ..., 4., 23., 23.],
[ 4., 23., 0., ..., 4., 23., 23.],
...,
[ 4., 4., 4., ..., 0., 4., 4.],
[ 4., 23., 23., ..., 4., 0., 23.],
[ 4., 23., 23., ..., 4., 23., 0.]])
import numpy as np
distortions = []
K = range(1, 8)
for k in K:
kmean_model = KMeans(n_clusters=k)
kmean_model.fit(adj_matrix)
dist = sum(np.min(cdist(adj_matrix, kmean_model.cluster_centers_, 'euclidean'), axis=1))
distortions.append(dist / adj_matrix.shape[0])
plt.figure(figsize=(10, 5))
plt.plot(K, distortions, 'bs-')
plt.xlabel('k')
plt.ylabel('Distortion')
plt.title('The Elbow Method showing the optimal k')
plt.show()
kmeans = KMeans(n_clusters=2)
kmeans.fit(adj_matrix)
kmeans_node_groups = [[] for _ in range(kmeans.n_clusters)]
for node, cluster in zip(G.nodes(), kmeans.labels_):
kmeans_node_groups[cluster].append(node)
print_communities(kmeans_node_groups)
plot_communities(G.nodes(), kmeans.labels_)
Number of communities: 2 Cluster 0: - 26 Nodes: As Seen on TV!, Baby, Basketball, Books, CDs, Cameras, Children's Clothing, Computers, Consumer Electronics, Crafts, DVDs, Garden, Golf Bags & Carts, Health and Beauty, Kids' Golf Clubs, Men's Clothing, Men's Golf Clubs, Music, Pet Supplies, Soccer, Sporting Goods, Strength Training, Toys, Video Games, Women's Clothing, Women's Golf Clubs - 0 Common values: Cluster 1: - 24 Nodes: Accessories, Baseball & Softball, Boxing & MMA, Camping & Hiking, Cardio Equipment, Cleats, Electronics, Fishing, Fitness Accessories, Girls' Apparel, Golf Apparel, Golf Balls, Golf Gloves, Golf Shoes, Hockey, Hunting & Shooting, Indoor/Outdoor Games, Lacrosse, Men's Footwear, Shop By Sport, Tennis & Racquet, Trade-In, Water Sports, Women's Apparel - 22 Common values: Canada, Caribbean, Central Africa, Central America, East Africa, East of USA, Eastern Asia, Eastern Europe, North Africa, Northern Europe, Oceania, South America, South Asia, South of USA, Southeast Asia, Southern Africa, Southern Europe, US Center, West Africa, West Asia, West of USA, Western Europe
from sklearn.mixture import GaussianMixture
n_clusters = 3
gmm = GaussianMixture(n_components=n_clusters)
gmm.fit(adj_matrix)
labels = gmm.predict(adj_matrix)
gmm_node_groups = [[] for _ in range(n_clusters)]
for node, cluster in zip(G.nodes(), labels):
gmm_node_groups[cluster].append(node)
print_communities(gmm_node_groups)
plot_communities(G.nodes(), labels)
Number of communities: 3 Cluster 0: - 24 Nodes: Accessories, Baseball & Softball, Boxing & MMA, Camping & Hiking, Cardio Equipment, Cleats, Electronics, Fishing, Fitness Accessories, Girls' Apparel, Golf Apparel, Golf Balls, Golf Gloves, Golf Shoes, Hockey, Hunting & Shooting, Indoor/Outdoor Games, Lacrosse, Men's Footwear, Shop By Sport, Tennis & Racquet, Trade-In, Water Sports, Women's Apparel - 22 Common values: Canada, Caribbean, Central Africa, Central America, East Africa, East of USA, Eastern Asia, Eastern Europe, North Africa, Northern Europe, Oceania, South America, South Asia, South of USA, Southeast Asia, Southern Africa, Southern Europe, US Center, West Africa, West Asia, West of USA, Western Europe Cluster 1: - 18 Nodes: Baby, Books, CDs, Cameras, Children's Clothing, Computers, Consumer Electronics, Crafts, DVDs, Garden, Health and Beauty, Men's Clothing, Music, Pet Supplies, Sporting Goods, Toys, Video Games, Women's Clothing - 3 Common values: Oceania, South Asia, Southeast Asia Cluster 2: - 8 Nodes: As Seen on TV!, Basketball, Golf Bags & Carts, Kids' Golf Clubs, Men's Golf Clubs, Soccer, Strength Training, Women's Golf Clubs - 3 Common values: Northern Europe, Southern Europe, Western Europe
def print_centrality(centrality, name):
result = pd.DataFrame(centrality.items(), columns=['Category Name', name])
result.sort_values(name, ascending=False, inplace=True)
print(result.to_records(index=False).tolist())
return result
def plot_centrality(centrality):
plt.figure(figsize=(15, 7))
node_color = [centrality[i] for i in centrality.keys()]
cmap = plt.cm.ScalarMappable(
cmap = 'Wistia',
norm = plt.Normalize(vmin=min(node_color), vmax=max(node_color))
)
cmap.set_array([])
plt.colorbar(cmap)
pos = nx.spring_layout(G)
nx.draw_networkx_nodes(G, pos, node_color=node_color, cmap='Wistia')
nx.draw_networkx_edges(G, pos, edge_color='grey', alpha=0.2)
plt.show()
degree = dict(nx.degree(G))
plot_centrality(degree)
print_centrality(degree, 'Degree Centrality').head()
[('Indoor/Outdoor Games', 49), ('Golf Balls', 49), ('Golf Shoes', 49), ('Crafts', 49), ('Golf Gloves', 49), ('Tennis & Racquet', 49), ('Fitness Accessories', 49), ('Cleats', 49), ("Children's Clothing", 49), ('Golf Apparel', 49), ('Lacrosse', 49), ('Baby', 49), ('Fishing', 49), ('Books', 49), ('DVDs', 49), ('CDs', 49), ('Accessories', 49), ("Girls' Apparel", 49), ('Hockey', 49), ('Shop By Sport', 49), ('Computers', 49), ('Cameras', 49), ('Consumer Electronics', 49), ('Camping & Hiking', 49), ("Men's Footwear", 49), ('Baseball & Softball', 49), ('Hunting & Shooting', 49), ('Water Sports', 49), ('Trade-In', 49), ('Cardio Equipment', 49), ('Boxing & MMA', 49), ('Electronics', 49), ("Women's Apparel", 49), ('Music', 41), ('Health and Beauty', 41), ('Toys', 41), ('Video Games', 41), ('Sporting Goods', 41), ('Pet Supplies', 41), ('Garden', 41), ("Men's Clothing", 41), ("Women's Clothing", 41), ('Strength Training', 40), ('Golf Bags & Carts', 40), ('Soccer', 40), ("Women's Golf Clubs", 40), ("Men's Golf Clubs", 40), ('Basketball', 40), ("Kids' Golf Clubs", 40), ('As Seen on TV!', 40)]
| Category Name | Degree Centrality | |
|---|---|---|
| 49 | Indoor/Outdoor Games | 49 |
| 26 | Golf Balls | 49 |
| 43 | Golf Shoes | 49 |
| 21 | Crafts | 49 |
| 41 | Golf Gloves | 49 |
betweenness = nx.betweenness_centrality(G)
plot_centrality(betweenness)
print_centrality(betweenness, 'Betweenness Centrality').head()
[('Indoor/Outdoor Games', 0.0018552875695732828), ("Girls' Apparel", 0.0018552875695732828), ('Golf Apparel', 0.0018552875695732828), ('Tennis & Racquet', 0.0018552875695732828), ('Fitness Accessories', 0.0018552875695732828), ('Cleats', 0.0018552875695732828), ('Golf Balls', 0.0018552875695732828), ('Golf Shoes', 0.0018552875695732828), ("Children's Clothing", 0.0018552875695732828), ('Lacrosse', 0.0018552875695732828), ('Baby', 0.0018552875695732828), ('Fishing', 0.0018552875695732828), ('Books', 0.0018552875695732828), ('DVDs', 0.0018552875695732828), ('CDs', 0.0018552875695732828), ('Hockey', 0.0018552875695732828), ('Golf Gloves', 0.0018552875695732828), ('Crafts', 0.0018552875695732828), ('Accessories', 0.0018552875695732828), ('Hunting & Shooting', 0.0018552875695732828), ('Electronics', 0.0018552875695732828), ('Shop By Sport', 0.0018552875695732828), ("Women's Apparel", 0.0018552875695732828), ('Computers', 0.0018552875695732828), ('Cameras', 0.0018552875695732828), ('Consumer Electronics', 0.0018552875695732828), ('Camping & Hiking', 0.0018552875695732828), ("Men's Footwear", 0.0018552875695732828), ('Baseball & Softball', 0.0018552875695732828), ('Water Sports', 0.0018552875695732828), ('Trade-In', 0.0018552875695732828), ('Cardio Equipment', 0.0018552875695732828), ('Boxing & MMA', 0.0018552875695732828), ('Golf Bags & Carts', 0.0), ('Music', 0.0), ("Women's Golf Clubs", 0.0), ('Toys', 0.0), ("Men's Golf Clubs", 0.0), ('Video Games', 0.0), ('Sporting Goods', 0.0), ('Health and Beauty', 0.0), ('Pet Supplies', 0.0), ('Garden', 0.0), ('Strength Training', 0.0), ("Men's Clothing", 0.0), ("Women's Clothing", 0.0), ('Soccer', 0.0), ('Basketball', 0.0), ("Kids' Golf Clubs", 0.0), ('As Seen on TV!', 0.0)]
| Category Name | Betweenness Centrality | |
|---|---|---|
| 49 | Indoor/Outdoor Games | 0.001855 |
| 18 | Girls' Apparel | 0.001855 |
| 44 | Golf Apparel | 0.001855 |
| 23 | Tennis & Racquet | 0.001855 |
| 24 | Fitness Accessories | 0.001855 |
closeness = nx.closeness_centrality(G)
plot_centrality(closeness)
print_centrality(closeness, 'Closeness Centrality').head()
[('Indoor/Outdoor Games', 1.0), ('Golf Balls', 1.0), ('Golf Shoes', 1.0), ('Crafts', 1.0), ('Golf Gloves', 1.0), ('Tennis & Racquet', 1.0), ('Fitness Accessories', 1.0), ('Cleats', 1.0), ("Children's Clothing", 1.0), ('Golf Apparel', 1.0), ('Lacrosse', 1.0), ('Baby', 1.0), ('Fishing', 1.0), ('Books', 1.0), ('DVDs', 1.0), ('CDs', 1.0), ('Accessories', 1.0), ("Girls' Apparel", 1.0), ('Hockey', 1.0), ('Shop By Sport', 1.0), ('Computers', 1.0), ('Cameras', 1.0), ('Consumer Electronics', 1.0), ('Camping & Hiking', 1.0), ("Men's Footwear", 1.0), ('Baseball & Softball', 1.0), ('Hunting & Shooting', 1.0), ('Water Sports', 1.0), ('Trade-In', 1.0), ('Cardio Equipment', 1.0), ('Boxing & MMA', 1.0), ('Electronics', 1.0), ("Women's Apparel", 1.0), ('Music', 0.8596491228070176), ('Health and Beauty', 0.8596491228070176), ('Toys', 0.8596491228070176), ('Video Games', 0.8596491228070176), ('Sporting Goods', 0.8596491228070176), ('Pet Supplies', 0.8596491228070176), ('Garden', 0.8596491228070176), ("Men's Clothing", 0.8596491228070176), ("Women's Clothing", 0.8596491228070176), ('Strength Training', 0.8448275862068966), ('Golf Bags & Carts', 0.8448275862068966), ('Soccer', 0.8448275862068966), ("Women's Golf Clubs", 0.8448275862068966), ("Men's Golf Clubs", 0.8448275862068966), ('Basketball', 0.8448275862068966), ("Kids' Golf Clubs", 0.8448275862068966), ('As Seen on TV!', 0.8448275862068966)]
| Category Name | Closeness Centrality | |
|---|---|---|
| 49 | Indoor/Outdoor Games | 1.0 |
| 26 | Golf Balls | 1.0 |
| 43 | Golf Shoes | 1.0 |
| 21 | Crafts | 1.0 |
| 41 | Golf Gloves | 1.0 |
eigenvector = nx.eigenvector_centrality(G)
plot_centrality(eigenvector)
print_centrality(eigenvector, 'Eigenvector Centrality').head()
[('Indoor/Outdoor Games', 0.14867064721396608), ("Children's Clothing", 0.14867064721396608), ('Crafts', 0.14867064721396608), ('Golf Gloves', 0.14867064721396608), ('Tennis & Racquet', 0.14867064721396608), ('Fitness Accessories', 0.14867064721396608), ('Cleats', 0.14867064721396608), ('Golf Balls', 0.14867064721396608), ('Lacrosse', 0.14867064721396608), ("Girls' Apparel", 0.14867064721396608), ('Baby', 0.14867064721396608), ('Fishing', 0.14867064721396608), ('Books', 0.14867064721396608), ('DVDs', 0.14867064721396608), ('CDs', 0.14867064721396608), ('Hockey', 0.14867064721396608), ('Golf Shoes', 0.14867064721396608), ('Accessories', 0.14867064721396608), ('Golf Apparel', 0.14867064721396608), ('Water Sports', 0.14867064721396608), ('Shop By Sport', 0.14867064721396608), ("Women's Apparel", 0.14867064721396608), ('Electronics', 0.14867064721396608), ('Boxing & MMA', 0.14867064721396608), ('Cardio Equipment', 0.14867064721396608), ('Trade-In', 0.14867064721396608), ('Hunting & Shooting', 0.14867064721396608), ('Baseball & Softball', 0.14867064721396608), ("Men's Footwear", 0.14867064721396608), ('Camping & Hiking', 0.14867064721396608), ('Consumer Electronics', 0.14867064721396608), ('Cameras', 0.14867064721396608), ('Computers', 0.14867064721396608), ('Video Games', 0.12767929833125463), ('Music', 0.12767929833125463), ('Garden', 0.12767929833125463), ('Toys', 0.12767929833125463), ('Pet Supplies', 0.12767929833125463), ('Health and Beauty', 0.12767929833125463), ('Sporting Goods', 0.1276792983312546), ("Men's Clothing", 0.1276792983312546), ("Women's Clothing", 0.1276792983312546), ('Strength Training', 0.12444087284685672), ('Golf Bags & Carts', 0.12444087284685672), ('Soccer', 0.12444087284685672), ("Women's Golf Clubs", 0.12444087284685672), ("Men's Golf Clubs", 0.12444087284685672), ('Basketball', 0.12444087284685672), ("Kids' Golf Clubs", 0.12444087284685672), ('As Seen on TV!', 0.12444087284685672)]
| Category Name | Eigenvector Centrality | |
|---|---|---|
| 49 | Indoor/Outdoor Games | 0.148671 |
| 28 | Children's Clothing | 0.148671 |
| 21 | Crafts | 0.148671 |
| 41 | Golf Gloves | 0.148671 |
| 23 | Tennis & Racquet | 0.148671 |
pagerank = nx.pagerank(G)
plot_centrality(pagerank)
print_centrality(pagerank, 'PageRank').head()
[('Indoor/Outdoor Games', 0.02910004330793566), ("Men's Footwear", 0.02910004330793566), ('Fishing', 0.02910004330793566), ('Lacrosse', 0.02910004330793566), ('Golf Balls', 0.02910004330793566), ('Cleats', 0.02910004330793566), ('Tennis & Racquet', 0.02910004330793566), ('Golf Apparel', 0.02910004330793566), ('Accessories', 0.02910004330793566), ('Camping & Hiking', 0.02910004330793566), ("Girls' Apparel", 0.02910004330793566), ('Cardio Equipment', 0.02910004330793566), ('Water Sports', 0.02910004330793566), ('Shop By Sport', 0.02910004330793566), ("Women's Apparel", 0.02910004330793566), ('Electronics', 0.02910004330793566), ('Boxing & MMA', 0.02910004330793566), ('Hunting & Shooting', 0.02910004330793566), ('Trade-In', 0.02910004330793566), ('Golf Gloves', 0.02910004330793566), ('Hockey', 0.028393028110741375), ('Golf Shoes', 0.028393028110741375), ('Fitness Accessories', 0.028393028110741375), ('Baseball & Softball', 0.028393028110741375), ('Crafts', 0.01460032971973412), ('Computers', 0.01460032971973412), ("Children's Clothing", 0.01460032971973412), ('Baby', 0.01460032971973412), ('Consumer Electronics', 0.01460032971973412), ('Books', 0.01460032971973412), ('DVDs', 0.01460032971973412), ('CDs', 0.01460032971973412), ('Cameras', 0.012939485390217484), ("Women's Golf Clubs", 0.01134315515704318), ("Men's Golf Clubs", 0.01134315515704318), ('Golf Bags & Carts', 0.01134315515704318), ('As Seen on TV!', 0.01134315515704318), ('Strength Training', 0.01134315515704318), ('Soccer', 0.01134315515704318), ("Kids' Golf Clubs", 0.01134315515704318), ('Pet Supplies', 0.009718050588254678), ('Health and Beauty', 0.009718050588254678), ('Music', 0.009718050588254678), ('Video Games', 0.009718050588254678), ('Garden', 0.009718050588254678), ("Men's Clothing", 0.009718050588254678), ("Women's Clothing", 0.009718050588254678), ('Toys', 0.009718050588254678), ('Sporting Goods', 0.009718050588254678), ('Basketball', 0.007820356856636094)]
| Category Name | PageRank | |
|---|---|---|
| 49 | Indoor/Outdoor Games | 0.0291 |
| 11 | Men's Footwear | 0.0291 |
| 31 | Fishing | 0.0291 |
| 29 | Lacrosse | 0.0291 |
| 26 | Golf Balls | 0.0291 |